CPU vs GPU rendering: which is best?
Time does not stand still, and as it often happens, what seemed premature and raw before is quite relevant today! It's exactly what can be said about GPU rendering.
The typical GTX 1070 8Gb gaming card is faster than the $ 3500 Intel Xeon 2699v4!
Due to the fact that the GPU is much more powerful than the CPU, such things as video editing, 3D rendering, as well as all kinds of mathematical and other resource-intensive calculations will be many times faster on graphic cards.
Many 3D rendering programs such as Cinema 4D, 3D max and others use GPU renderers like VrayRT 3.5+, OTOY Octane, Corona, Fstorm and others. Rhino and Maxwell render also started to be friends with the GPU. This approach allows you to significantly speed up the rendering time.
Many famous studios use their own developments. For example, Pixar Renderman for Blender. And Pixar offers Renderman for free personal use. Video encoding methods do not stand still - new codecs for Sony Vegas (MAGIX AVC / AAC MP4) are available, and Pinnacle Studio and the popular After Effects CC have long been able to use GPU resources. CAD applications also do not stand aside.
But it doesn't end there! You can even enable GPU hardware acceleration in Corel VideoStudio Pro and After Effects CC. Many new codecs with very good rendering capabilities in 4K or even 8K HDR, sharpened for GPUs. For instance, Daniel2 codec.
What to use for the GPU?
For example, Nvidia RTX 2070 / 80Ti accelerators, as well as the GTX 1060 series and above, are great for GPU rendering. The AMD VEGA accelerators and the RADEON RX series are slightly less popular, although they can be quite useful as well.
The benchmark is Nvidia QUADRO cards and Nvidia VCA stations. Just prepare many thousands of dollars for this. You can find an analog that will be cheaper.
But there is a very important point. The amount of video memory in a GPU is very small compared to expensive systems. Usually 8-16GB, and this may not be enough for everyone. Moreover, when using several accelerators in a GPU rendering, the video memory is NOT summed up!
For studios that produce a large number of renders, power consumption is another aspect to consider.
Surprisingly, despite the large difference in price between CPU and GPU, performance per watt is strikingly similar for devices with a large number of cores.
The GTX 1070 and Xeon 2699v4 both have a peak power consumption of around 150W and perform roughly the same. Therefore, no matter what device you use, the hardware of the same generation should use approximately the same amount of power.
However, low-core, high-frequency processors such as the Intel 7700k tend to consume more power than low-speed multi-core processors.